# HarnessRL: Co-Evolving Policies with a Just-in-Time Training Harness for Open-Ended RL > Source: > Published: 2026-10-21 05:30:02+00:00 Rubric-based reinforcement learning (RL) trains policies for open-ended generation with natural-language criteria as rewards, but these rewards can become ineffective as training progresses. We identify two failure modes. Under non-discriminative rewards, scores collapse to uniformly low or high val